A Wavelet-Based Outlier Detection and Noise Component Analysis for GNSS Position Time Series

Abstract

Various signals of crustal deformation and mass loading deformation are contained in a GNSS position time series. However, a GNSS position time series is also polluted by outliers and various colored noise, which must be reasonably modelled before estimating deformation signals. Since temporal signals of the GNSS position time series are non-linear and complicated, we propose a wavelet-based approach for outlier detection, which first retrieves the temporal signals from the GNSS position time series by using wavelet analysis, and then detect outliers in the residual position time series by using the interquartile range. After the detected outliers are eliminated from the residual time series, the noise components, including white noise and flicker noise, are estimated by using MINQUE approach. Our proposed approach is used to process the real GNSS position time series of the Crustal Movement Observation Network of China (CMONOC) over the period spanning 1999–2018. The results demonstrate that our approach can detect the outliers more efficiently than the traditional approach, which retrieves the temporal signals by using a functional model with trend and periodic variations. As a result, the noise components estimated with our proposed approach are smaller than those with the traditional approach for the GNSS position time series of all CMONOC stations.